Pico W Wifi Troubles #16471

edmcman · 2024-12-22T16:30:27Z

edmcman
Dec 22, 2024

I'm having trouble with the Pico W connecting to Wifi. It appears that if it is unused for a while, the Pico disconnects but can't seem to detect this.

I get weird debug messages when this happens:

[ 5322065] ASYNC(0000,49,0,0,0)

If I make a request every 5 minutes, the problem does not happen.

When the Pico disconnects, network.isconnected() returns True and network.status() returns 3.

Any idea what is going on? Is my pico just flaky perhaps?

edmcman · 2024-12-22T16:48:54Z

edmcman
Dec 22, 2024
Author

Current code located here https://github.com/edmcman/rpi-temp-server/tree/debug

4 replies

mendenm Dec 22, 2024

My solution to this was to install uping, and just ping the network every so often. It's a hack, but made connections stable for months.

davefes Dec 22, 2024

I was interested in your solution so checked your GitHub project. I don't see anything in main.py regarding uping. How do you get it to work?

Curious about the sequence of events:

but made connections stable for months.

Have you used this before on other projects?

mendenm Dec 22, 2024

here is the main loop I used for a powerline phase monitor, along with ancillary code. The keep_net_alive() function is what you may be interested in. It's not pretty, but it works.

#test microdot

from microdot_asyncio import Microdot
import asyncio
import uping

app = Microdot()

@app.get('/')
async def index(request):
    #print(request.client_addr, request.body)
    return '''
<html>
<head> </head>
<body>
<h1>Hello, World!</h1>
</body>
</html>
''', 200, {'Content-Type': 'text/html'}


@app.get('/set_time')
async def set_time(request):
    #print(request.client_addr, request.body)
    import ntptime
    ntptime.timeout = 10
    ntptime.settime()
    from machine import RTC
    rtc = RTC()
    return f'''
<html>
<head> </head>
<body>
<h1>{str(rtc.datetime()):s} UTC</h1>
</body>
</html>
''', 200, {'Content-Type': 'text/html'}


@app.get('/index/')
def index_files(request):
    import os
    top = '''
<html>
<head> </head>
<body>
<ol>
'''
    vals = '\n'.join( "<li>"+x+"</li>" for x in sorted(os.listdir("/timelogs/")) )
    
    bot = '''
</ol>
</body>
</html>
'''
    return top + vals + bot, 200, {"Content-type":"text/html"}

@app.get('/timelogs/<string:filename>')
def handle_file(request, filename = None):
    from microdot_asyncio import send_file
    args = request.args
    delete = 'delete' in args and args['delete'][0] == '1'
            
    try:
        if delete:
            import os
            os.remove('/timelogs/' + filename)
            return 'removed', 200, { 'Content-Type':'text/plain' }
        else:
            return send_file('/timelogs/' + filename, content_type = 'text/plain',
                 extra_headers = {'Content-Disposition':  'attachment'}
                         )
    except OSError:
        return "File not found:" + filename, 404, {}
    
from detect_power_line import doit_events, current_status

@app.get('/status')
def index_files(request):
    import os
    from machine import RTC
    rtc = RTC()
    
    top = '''
<html>
<meta charset="UTF-8">
<style>
table {
  border-collapse: collapse;
}
th, td {
  padding-right: 5px;
  padding-left: 5px;
  padding-top: 5px;
}
th {
  text-align: center;
  border: 2px solid black;
}
td {
  text-align: right;
  border: 1px solid black;
}
</style>
'''
    head = '''
<head> Date: {0:04d}-{1:02d}-{2:02d} {4:02d}:{5:02d}:{6:02d}</head>
'''.format(*rtc.datetime())

    body = '''
<body>
<table>

<tr>
<th> time / ms </th>
<th> ADC </th>
<th> IQ mag / mLSB </th>
<th> I </th> <th> Q </th> <th> φ/° </th>
<th> RMS / LSB </th>
<th> Δt / s </th>
</tr>
'''
    
    bot = '''
<table>
</body>
</html>
'''

    vals = '\n'.join(
        [ '\n'.join(['<tr>'] + ["<td>" + x+ "</td>" for x in r.split()] + ['</tr>'])  for r in current_status ]
    )
    
    return top + head + body + vals + bot, 200, {"Content-type":"text/html", "Refresh":"10"}

nic = None

netdict = json.load(open("networks.json","r"))

async def keep_net_alive(verbose):
    global nic
    import network
    last_status = -1
    while nic != "STOP":
        stat = nic.status() if nic is not None else -4
        if stat < 0:
            if verbose and nic is not None: print("no net:", nic.status(), nic.ifconfig())
            if nic is not None: nic.active(False)
            last_status = stat
            #network hard crashed...
            network.hostname("powerline_pico")
            network.country('US')
            #ap = network.WLAN(network.AP_IF)
            #ap.active(False) #turn off any AP interface activity
            nic = network.WLAN(network.STA_IF)
            nic.active(True)
            await asyncio.sleep(1)
            if verbose: print("attempted activation:", nic.status())
        elif last_status < 0 or not nic.isconnected(): #alive, but not connected
            nets = nic.scan()
            if verbose: print("found", [str(x[0],'utf8') for x in nets])
            for net in nets:
                netname = str(net[0],"utf8") #might be bytes from NIC
                if netname in netdict:
                    nic.disconnect() #start  over here
                    nic.active(False)
                    await asyncio.sleep(1)
                    nic.active(True)
                    nic.connect(net[0], netdict[netname])
                    while nic.status() > network.STAT_IDLE and nic.status() != network.STAT_GOT_IP: 
                        await asyncio.sleep(1) #check again later
                    if verbose: print("connected to", str(net[0], 'utf8'), nic.ifconfig(), nic.config('hostname'))
                    break

            nic.config(hostname  = "powerline_pico")
            last_status = nic.status()
            await asyncio.sleep(1) #check again later
        else:
            config =  nic.ifconfig()
            uping.ping(config[-1], count = 1)
            if verbose: print("on net:", nic.status(), nic.ifconfig())
            await asyncio.sleep(5)            

async def set_initial_time(log):
    import ntptime
    while 1:
        try:
            if log: print("setting time")
            ntptime.timeout = 5
            ntptime.settime()
            if log:
                from machine import RTC
                rtc = RTC()
                print("set time to:", rtc.datetime())
            break
        except OSError as e:
            print("oops", str(e))
            await asyncio.sleep(5)
            pass #just keep looping until it happens
    
async def doit(log = False, startidx = 0):
    global nic
    #print("starting line monitor")
    net_monitor = asyncio.create_task(keep_net_alive(log))
    
    time_init = asyncio.create_task(set_initial_time(log)) #should return after time set, when network is up
    
    line_monitor = asyncio.create_task(doit_events(log = log))
    #print("starting web  server")
    web_server = asyncio.create_task(app.start_server(host='0.0.0.0', port=5000, debug = False, ssl = None))
    await line_monitor
    app.shutdown()
    nic = "STOP"
    await net_monitor
    
if __name__=="__main__":
    asyncio.run(doit(log=True))

davefes Dec 22, 2024

Thanks. I will have a look at that as it could eliminate a ping client and watchdogs on my existing server.

edmcman · 2024-12-22T20:08:39Z

edmcman
Dec 22, 2024
Author

I'm running this fish script: set sleep 1; while true; date; echo "Sleeping $sleep"; sleep $s leep; wget -q http://192.168.1.4/temp -O /dev/stdout || break; set sleep (math "2*$sleep"); end ; not ify done

It's currently past 4096 and sleeping 8192.

Hard to say because it is non-deterministic, but I did just add wlan.config(pm=network.WLAN.PM_NONE). Maybe that's the solution. Which still feels like a bug.

5 replies

davefes Dec 22, 2024

Note: I only have ESP32 WiFi experience.

Some other hints found floating around:

config(timeout_ms=xx), PM_NONE and txpower (set to some value)

The only way I have been able to maintain a reliable connection for an alarm server to a hotspot dongle is to run a ping client and if the server doesn't process the ping within a certain time it re-boots. Likewise, if the ping client doesn't receive an ack, then it will re-boot as well.

You then have to handle re-connects properly.

edmcman Dec 22, 2024
Author

So with PM_PERFORMANCE it failed after 1024 seconds. Trying again with PM_NONE.

It's sad to hear how many people have had to resort to ping hacks!

edmcman Dec 22, 2024
Author

On PM_NONE, I'm at 4096 seconds again. I'll try to let it run some more.

Out of curiosity, has anyone else that has faced this problem tried PM_NONE?

davefes Dec 22, 2024

The two guys that have helped in the past are:
glenn20 and shariltumin
I recall they both mentioned `PM_NONE'

That suggestion happened after I got the ping client setup working and couldn't be bothered, at the time, to try it.

I usually takes months for connect failure so I got tired of testing.

hippy on the Pico Micropython forum might be worthwhile searching or having a chat with.

Good luck!

davefes Dec 22, 2024

Here is my latest wifi_functions.py script, tested but not been used at the moment:

import network
from time import sleep
import config


network.country("NZ")   # New Zealand
sta_if = network.WLAN(network.STA_IF)

sta_if.active(False)  # TOGGLING THIS FIRST PREVENTS "OSError: Wifi Internal Error"
sta_if.active(True)

#sta_if.config(pm=sta_if.PM_NONE)  # disable power management, seems to need active(TRUE) first
sta_if.config(pm=sta_if.PM_PERFORMANCE)   # this works fine for me, and saves power/heat

sta_if.config(txpower=10)  # Fixes wifi ESP32 C3 5x5 WS2812 boards
sta_if.connect(config.hotspot, config.password")

print('Connection pending')

while not sta_if.isconnected():
    if sta_if.status() == 1000:
        print('Waiting... status: {}...'.format(sta_if.status()))
    else: 
        print('Status: {}...'.format(sta_if.status()))

    sleep(1)
   
print('Completed with {}...'.format(sta_if.status()))  # 1010 is connected.

if sta_if.status() == 1010:
    print('Wifi connection successful')

edmcman · 2024-12-22T22:24:45Z

edmcman
Dec 22, 2024
Author

Thanks all. I also just found #9455 which seems like the same problem. I will post more about the issue there.

2 replies

scruss Dec 23, 2024
Sponsor

There's been a lot of work done on network_cyw43 since that issue was posted, so it might not be best to revisit that one

edmcman Dec 24, 2024
Author

Thanks, I'll make a new issue when I am ready then. I'm running tests and collecting data right now.

edmcman · 2024-12-24T17:47:25Z

edmcman
Dec 24, 2024
Author

I'm still continuing to test. I am less sure that the PM fixes the issue. I was testing using a laptop, and I think a significant factor may have been moving or flipping the pico. I am moving to an isolated computer so this won't be an issue anymore!

4 replies

scruss Dec 24, 2024
Sponsor

Maybe expecting a wifi micro-controller to be always available on the network is a bit of a stretch. The MQTT model of the microcontroller (client) waking up periodically, taking a reading, connecting to the network and sending that reading to a broker is pretty robust. Peter Hinch's MicroPython Asynchronous MQTT is remarkably resilient.

But MQTT is of no use if you need to query your sensor right now

edmcman Dec 24, 2024
Author

Maybe expecting a wifi micro-controller to be always available on the network is a bit of a stretch.

What do you mean by this? I'm not expecting 100% availability, but I do expect that if the micro-controller gets knocked offline or drops a few packets, it is able to recover. Just like most devices that are on wifi.

What is happening now is some unknown cause -- my current guess is maybe an AP disconnect or dropped packets -- causes the micro-controller to enter a state where it disconnects from the wifi but reports that it is still connected to micropython. It never recovers. I consider this a bug (and unacceptable, even for a micro-controller). Hopefully that's not under dispute!

scruss Dec 24, 2024
Sponsor

I'm not disputing that falling off a network yet still reporting connected is unacceptable. It's really annoying.

What I was suggesting was that — in my experience — networked micro-controllers haven't typically been used in an always-on/always-available configuration. They've typically taken a measurement, connected to a network, sent the measurement, then disconnected. This is a fairly robust model, as the connection is opened then data is sent almost immediately.

I don't know what a Pico W does in the background to keep its connection open. Clearly, not enough to stay connected to some APs

edmcman Dec 25, 2024
Author

My son also suggested a similar approach when I told him about the problem. This is a personal project for me, and I would like to find out and hopefully fix whatever is going wrong so that others don't encounter the same problem. If that proves untenable, I might switch to storing measurements in a more stable location.

edmcman · 2024-12-25T14:13:39Z

edmcman
Dec 25, 2024
Author

I've been running experiments for almost a day now, with the power management choice selected randomly from: ["None", "0xa11140", "network.WLAN.PM_NONE", "network.WLAN.PM_PERFORMANCE", "network.WLAN.PM_POWERSAVE"]

I limited each test to intervals of no more than 2048 seconds. I will probably try longer intervals later.

We are currently on the 18th test. Out of the 17 that completed, 8 failed. Here is the breakdown of PMs for the failures:

  6 network.WLAN.PM_POWERSAVE
  2 None

But overall the PMs aren't quite well distributed yet:

  3 0xa11140
  6 network.WLAN.PM_NONE
  7 network.WLAN.PM_POWERSAVE
  3 None

But I think for me the big take away is that the problem does occur with None (the default PM mode). It is too early to say for sure, but 0xa11140 and PM_NONE might fix or make the problem less likely.

1 reply

edmcman Dec 25, 2024
Author

The None bit is interesting because I believe micropython adopted the default specifically because it was more reliable, though I don't have the issue link handy now.

edmcman · 2024-12-25T19:30:21Z

edmcman
Dec 25, 2024
Author

Since I have enough data to make a coherent issue, I did so here: #16482

I will close this discussion in favor of this issue. Thank you all for the input and help so far.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

MicroPython

Pico W Wifi Troubles #16471

{{title}}

Replies: 6 comments 16 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

Pico W Wifi Troubles #16471

Replies: 6 comments · 16 replies

edmcman Dec 22, 2024 Author

edmcman Dec 22, 2024 Author

edmcman Dec 22, 2024 Author

edmcman Dec 22, 2024 Author

edmcman Dec 22, 2024 Author

scruss Dec 23, 2024 Sponsor

edmcman Dec 24, 2024 Author

edmcman Dec 24, 2024 Author

scruss Dec 24, 2024 Sponsor

edmcman Dec 24, 2024 Author

scruss Dec 24, 2024 Sponsor

edmcman Dec 25, 2024 Author

edmcman Dec 25, 2024 Author

edmcman Dec 25, 2024 Author

edmcman Dec 25, 2024 Author

Replies: 6 comments 16 replies

edmcman
Dec 22, 2024
Author

edmcman
Dec 22, 2024
Author

edmcman Dec 22, 2024
Author

edmcman Dec 22, 2024
Author

edmcman
Dec 22, 2024
Author

scruss Dec 23, 2024
Sponsor

edmcman Dec 24, 2024
Author

edmcman
Dec 24, 2024
Author

scruss Dec 24, 2024
Sponsor

edmcman Dec 24, 2024
Author

scruss Dec 24, 2024
Sponsor

edmcman Dec 25, 2024
Author

edmcman
Dec 25, 2024
Author

edmcman Dec 25, 2024
Author

edmcman
Dec 25, 2024
Author